NeuroFusion: Multimodal neuroimage fusion in pre-neurosurgical evaluation

Thesis type: 



Nádia Moreira da Silva


João Paulo Trigueiros da Silva Cunha


NeuroFusion: Multimodal neuroimage fusion in pre-neurosurgical evaluation


The success of neurosurgery strongly depends on the pre-neurosurgical evaluation phase, in which the delineation of the areas to be removed or to be stimulated must be precise.
In the case of Epilepsy, the interpretation of the iEEG data can be enhanced by an exact detection of the position of the electrodes on the cortex. Therefore, the epileptogenic foci and the eloquent areas can be more accurately delineated. For patients undergoing deep brain stimulation, the delineation of the target areas prior to surgery and after the implantation is fundamental, as well as the confirmation of the electrodes implanted in these exact areas. Improvements in the identification of target areas and in electrodes positioning leads to a successful surgery and consequently improve the patient’s outcome and quality of life.

For this study a collaboration was established with Dr. Ricardo Rego from Hospital São João, for the Epilepsy cases, and with Dr. Verena Rozanski from Munich University Hospital, for the patients with Parkinson and Dystonia, undergoing deep brain stimulation.

A pipeline and an interface were developed to accurately detect the subdural and deep brain electrodes, respectively. The positions of the deep brain electrodes were compared with the ones given by Dr. Verena Rozanski and differences of less than a voxel dimension were observed. Thus, the interface developed can be widely used to produce automatically the electrodes masks. The subdural electrodes were also accurately segmented without residues of skull, artefacts or even beam hardening. In order to enhance the visualization of the strips and grids over the cortex, cerebellum was removed.

Our tool was used in 3 iEEG epileptic patients and in the last one our results were part of the surgery decision procedure. A 3D model of MRI dataset, without cerebellum, overlaid with the subdural electrodes mask was created using MRIcron. The 3D model was used by HSJ for the pre-neurosurgical evaluation. The patient in which this approach was presurgically applied has being seizure-free since surgery, performed one month ago. More requests has been made by HSJ for future patients.
For the segmentation of the target areas for deep brain stimulation and others in the deep brain area, a recent automatic method available in FSL was used. The execution time of the automatic segmentation process, for each structure, was less than two minutes. The resulting structures masks were very congruent in shape and position with the corresponding area in the MRI dataset from the patient. Furthermore, the results obtained allows us to evaluate the performance and improve our knowledge of this recent method and therefore estimate their potential in future applications.